Mining Opinion Leaders Of Chinese Microblogs Based On Multi-Objective Optimization Algorithm Within The Clould Computing Framework

In recent years, with the popularity of the Internet, the way people communicate has undergone enormous changes. As a new network information transmission media, microblog is popular in the majority of Internet users. It allows the information spread widely in a very short time with the timeliness, so it is very important to control the information effectively. For if someone deliberately spreads false information or publishes remarks of endangering social security, the consequences would be disastrous. So it is significant of mining and monitoring opinion leaders who dominates the network information dissemination.Currently mining method for opinion leaders are mainly statistical methods, cluster analysis, social network analysis based on SNA method, etc. These methods all have their own features. But they cannot exhibit good processing capabilities facing with the massive microblog datas produced by331million microblog users.In this paper, attributes of microblog users are treated as a foothold. They are combined with multi-objective optimization problem, and skyline query is introduced into the mining opinion leaders in microblog. is a kind of method to solve the multi-objective optimization problem. Facing the massive microblog data, the MapReduce framework that is the key technology of Hadoop is introduced in this paper. The block nested loop algorithm and sort filter algorithm of skyline are realized in this programming framework. They have better performance when processing huge amounts of data. Then the evaluation model of microblog opinion leader is established. Two indicators are used to evaluation microblog opinion leaders, they are user influence and the degree of user involvement. At last using AHP to determine the weight of each attribute and the formula for mining opinion leaders is given. In the period of mining experiments, this paper builds Hadoop cluster environment, and designs microblog crawlers to obtain the datas. Then the datas obtained by crawlers are processed with the SFS algorithm, and are calculated within the evaluation model of microblog opinion leader. At last, the opinion leaders mined in this paper are compared with official popular users of Sina Weibo. The result shows that the opinion leaders we found adopted the Block Nested Loops algorithm had a wide spreading existence in social fields, otherwise it focused on entertainment in the Sina Weibo. Therefore, the methodology of this research provides an effective solution for handling the scenario with large amount and high dimensionality of data, and can be adopted to find opinion leaders in a way.